Synthesis of optimal control systems: A comparison between model checking and dynamic programming techniques

Abstract

In the last years, the use of sophisticated controllers, i.e., components that control larger systems, has become very common. In previous papers, we introduced a general, automatic methodology for the synthesis of optimal controllers which exploits explicit model checking to generate controllers for very complex systems. However, to completely validate the usefulness of our methodology we still need to compare it with other, well-established controller generation techniques that are currently applied to a variety less complex control problems. To this aim, in this paper we present a comparative evaluation, from both a theoretical and a computational point of view, between our model checking methodology and the widely applied dynamic programming approach, using the well known inverted pendulum problem as a test case. As a result, we find that our approach is always feasible when dynamic programming can be applied, and that the generated controllers perform better than the ones obtained by dynamic programming. © Springer Science+Business Media B.V. 2008.